Problems with inward PPTP tunnel

[Brett Glass]

At 05:31 AM 3/24/2009, Luiz Otavio O Souza wrote:

>>Hi,
>>
>>I read over your patch and I don't think you need to
>>change the definition of rt_Update() to fix this bug.
>>
>>Have you read my patch committed to head/user.sbin/ppp/route.c
>>on 12/18/2008?  svn r186308, cvs r1.96.
>>
>>Please take a look.
>>
>>Thanks,
>>
>>-- Qing
>
>Hi Qing,
>
>Yes i've read your patch, but i don't understand what you are 
>meaning... and yes, changing the definition of rt_Update is not my 
>first intention, but it is the way i've found to fix this.
>
>Backing to the patch... The rt_Update need the ifp and ifa 
>information to correctly update the route, and this is available 
>only in route_UpdateMTU (wich read the current route table).
>
>You are suggesting that this information could be found at 
>sa[RTAX_GATEWAY] (if sa[RTAX_GATEWAY]->sa_family == AF_LINK) ? And 
>i don't need to pass the sa[RTAX_IFP] and sa[RTAX_IFA] ?
>
>Well, i really don't remember, but i'll check and post an update later.
>
>Thanks,
>Luiz

I personally think that, from a style perspective, changing the 
definition of rt_Update is good. I was thinking of doing it myself 
when I was inspecting the code and trying to figure out how to 
ensure that the interface was specified in the routing table entry. 
Adding the additional arguments makes the routine more flexible and 
makes the intent of calls to it more explicit. That being said, any 
change that fixes the bug is welcome. (The code that shipped with 
7.1-RELEASE will break many if not most systems that rely on 
userland PPP to implement a server.) I might consider using mpd 
instead, but it would be a big implementation change and mpd lacks 
a few capabilities that I really need.

--Brett Glass